TikTok boosts Mamdani videos, suppresses Cuomo content: study

Yehonatan Dodeles, the CTO of Spring AI, wrote on Medium that early analysis has suggested the algorithm findings, "a pattern that could meaningfully influence public perception and voter behavior." He said the preliminary findings were being shared now "because the issue is timely and urgent. Early evidence points to algorithmic influence that may be shaping voter perception in the New York elections."

Dodeles explained that Spring AI, "Based on deep reverse-engineering of TikTok’s recommendation system and representative data (real & synthetic users)," built an AI model "that measures when videos get more exposure than they should. We call these 'Excessive Publicity' videos." It was found that around 17 percent of all videos were labeled as having "Excessive Publicity," and the model "correctly flagged" 76.4 percent of videos labeled as ads as "promoted," "confirming it works."

Spring AI looked at keywords to see if anything deviated from the 17 percent standard publicity established by the study and found one that "exploded off the chart: Politics," with 55 percent of political videos getting excessive promotion.

It was found that among political keywords, Mamdani and Cuomo’s names had the highest percentages of videos receiving excessive publicity, at 74 and 79 percent, respectively.

"Content favoring Mamdani, both videos supporting him and those opposing Cuomo, received significantly higher rates of “Excessive Publicity” compared to content favoring Cuomo. While both sides showed evidence of non-organic promotion, the disparity is clear: pro-Mamdani content consistently received more amplification than pro-Cuomo content," Dodeles wrote.

"To isolate this disparity within the political space, we normalized our measurements against the 55% political baseline. Pro-Mamdani content substantially exceeded this already-elevated baseline, while pro-Cuomo content fell below it — suggesting not just relative disadvantage but active suppression," he added.

Arguing against those who may suggest that the pattern reflects TikTok’s younger demographic favoring the socialist candidate, Dodeles wrote, "that misses the point. We’re not measuring how much content exists for each candidate, or even how popular that content is. We’re measuring deviation from expected organic reach."

Dodeles wrote that it’s "not just bias," it’s "algorithmic steering." He stressed that the findings are preliminary and more research should be done, but "the implications are huge."